An LLM’s not going to tell you no

The advantage of working with a UX professional is that he or she will challenge ideas and work through the tradeoffs of complex problems to find a solution. In most cases the UX designer’s solution won’t be the simplistic “just put another button there” that management asked for. That’s the job, though. You unpack iffy suggested solutions to find the real problem, and then you work from that to flesh out a better solution.

The issue with using an LLM is that it will do precisely what it’s told. It will give managers the button and text they want, but it will never tell them that an email to select customers would be a better way to communicate that information.

I’ve put together my own amateur anthropologist’s take on this. Almost every person I know who’s “excited about AI” is a man. And a certain type at that. And for them, there’s something thrilling about having an entity that can’t say no to you, that can’t reject you. That’s why I don’t think the concern about chatbot “girlfriends” is aimless moral panic. This technology is reinforcing the fantasy of those already in power: they should never be told no to and need their every whim indulged.

As an aside and following the lead of Jaron Lanier, I refuse to say AI. It’s just a bunch of computation mashing up text. There’s no intelligence, no non-human thinking going on behind the scenes.

That’s why I’ve come to both crave and value human craft and thought. There’s so much joy in reading a beautiful novel that a person has put his heart and soul into drafting. Using well-designed software is a pleasure. And yet more and more of the digital world is becoming some sort of LLM dump, to borrow from the idea that the internet is an SEO landfill. Within a few years, I suspect that the internet will become mostly useless for finding anything other than the Current Thing and LLM generated content.

No thanks. I’m enjoying spending a lot less time online these days.